Further methods for model fitting and inference

The fitR package introduced in this course was developed as a teaching tool, and the implementation of inference algorithms in this package is not particularly efficient of stable. For vastly superior alternative options for fitting models to data there are much better options. Here we provide some additional short tutorials that show how these could be used to fit the SEITL and related models to the Tristan da Cunha data set.

• Fitting models to data using pomp. For more information, see also the tutorials at the pomp web site.
• The SEITL model in LibBi. For more information see also the LibBi web site.
• We also recommend looking at stan a general probabilistic programming language for Bayesian inference, as well as the great tutorials for using them with infectious disease published by Chatzilena et al. (2019) and Grinsztajn et al.
• The mcstate package for model fitting, combined with odin, dust and odin.dust is implements several of the methods introduced in this course. See the vignette on fitting SIR models for an introductory example.
• For a more general discussion of different approaches to fitting mechanistic epidemic models to data including a comparison of different approaches to modelling the observation process, make sure you read Fitting mechanistic epidemic models to data: A comparison of simple Markov chain Monte Carlo approaches by Li et al.

 

---

This web site and the material contained in it were originally created in support of an annual short course on Model Fitting and Inference for Infectious Disease Dynamics at the London School of Hygiene & Tropical Medicine. All material is under a MIT license. Please report any issues or suggestions for improvement on the corresponding GitHub issue tracker. We are always keen to hear about any uses of the material here, so please do get in touch using the Discussion board if you have any questions or ideas, or if you find the material here useful or use it in your own teaching.